Abstract
As a part of the possible technology to achieve a method to remove weeds physically without using any herbicide, a fuzzy inference system was developed. The system recognizes crop projections on a binary image, where both crop projections and noises coexist, taking advantage of the grid-like regularity of the crop distribution on the field. As a result, the crop recognition accuracy was improved, and the precondition that “target crops must be larger than weeds” is no longer necessary for the discrimination of crops and weeds.
